From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from [140.186.70.92] (port=51563 helo=eggs.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1Pmsgq-0005t8-45
	for qemu-devel@nongnu.org; Tue, 08 Feb 2011 13:53:30 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <peter.maydell@linaro.org>) id 1Pmsgl-0004EE-BP
	for qemu-devel@nongnu.org; Tue, 08 Feb 2011 13:53:24 -0500
Received: from mail-ew0-f45.google.com ([209.85.215.45]:35084)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <peter.maydell@linaro.org>) id 1Pmsgk-0004Dx-VF
	for qemu-devel@nongnu.org; Tue, 08 Feb 2011 13:53:23 -0500
Received: by ewy10 with SMTP id 10so3256953ewy.4
	for <qemu-devel@nongnu.org>; Tue, 08 Feb 2011 10:53:21 -0800 (PST)
MIME-Version: 1.0
In-Reply-To: <4D5182BF.9050002@st.com>
References: <4D50252A.2000100@st.com>
	<AANLkTim_TSvJVnyHFzmOiu52bmMu3_d=4Uz690S15o=1@mail.gmail.com>
	<4D5182BF.9050002@st.com>
Date: Tue, 8 Feb 2011 18:53:21 +0000
Message-ID: <AANLkTi=uDtGamByjJ3ZA0uMHU446N3=01bTPrgXmeLTM@mail.gmail.com>
Subject: Re: [Qemu-devel] [PATCH v3] Softfloat: Add support to softfloat to
	return floatxx_default_nan when,
	the corresponding target status flag is set.
From: Peter Maydell <peter.maydell@linaro.org>
Content-Type: text/plain; charset=UTF-8
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Christophe Lyon <christophe.lyon@st.com>
Cc: "qemu-devel@nongnu.org" <qemu-devel@nongnu.org>

On 8 February 2011 17:51, Christophe Lyon <christophe.lyon@st.com> wrote:
> On 08.02.2011 16:59, Peter Maydell wrote:
>> This is float32_to_float16() but it looks like you've missed
>> float16_to_float32() ?
>
> Well, actually I somewhat managed to persuade myself that it was not necessary to patch float16_to_float32(), but looking at it again it seems there are more bugs in that function:
> - the ieee parameter shouldn't be used (my ARM ARM does not mention a different behaviour when AH bit is set)

In the ARM ARM pseudocode, FPUnpack() of a 16 bit value looks
at the FPSCR AH bit, so float16_to_float32() does need to depend
on the ieee flag.

> - it calls float32ToCommonNaN(a) where 'a' is a float16 instead of a float32

It's only doing that to raise the exception, but you're right that
it ought to do things properly...

> - this means implementing float16ToCommonNaN, thus float16_is_signaling_nan()

...like that, yes.

I'm just working through a set of bugs in float32_to_float16
and float16_to_float32 where they generally mishandle
various things.

I have also just noticed this error in your patch:

+static bits16 commonNaNToFloat16( commonNaNT a STATUS_PARAM)
+{
+    bits16 mantissa = a.high>>57;

That should be ">> 54". (We're shifting a mantissa which is
in a 64 bit value down so that the top bits of it end up in
the bottom 10 bits (corresponding to the size of mantissa
on a float16); so we want to shift by 64-10 == 54.)

-- PMM